Systems and methods for assistive document retrieval in data-sparse environments

Abstract

Methods and systems are described herein for novel uses and/or improvements to artificial intelligence applications for data-sparse environments. As one example, methods and systems are described herein for overcoming the problem of determining an appropriate digital asset to recommend to a user in real-time based on training data that may be shared between multiple intents (e.g., as would be found in responses to textual, verbal, and/or other real-time communications). In particular, the methods and systems overcome this technical problem by using an artificial intelligence trained to identify specific verbiage of a user to display digital assets corresponding to a user's intent thereby increasing productivity, organization, and collaboration, and reducing frustration.

What is claimed is: 
     
         1 . A system for generating digital asset recommendations in real-time communication applications using artificial intelligence models in data-sparse environments, the system comprising:
 one or more processors; and   a non-transitory, computer-readable medium comprising instructions that when executed by the one or more processors cause operations comprising:
 receiving, at a user interface of a real-time communication application of a mobile device, a first action, wherein the first action comprises a phrase entered by a user; 
 determining a first subset of the phrase, wherein the first subset comprises one or more alphanumeric characters; 
 determining whether the first subset corresponds to a first labeled subset in a plurality of labeled subsets, wherein each labeled subset of the plurality of labeled subsets corresponds to one of respective known intents of the user, and wherein the plurality of labeled subsets is based on an artificial intelligence model that is trained to:
 determine that a respective labeled subset of the plurality of labeled subsets was entered by the user into the real-time communication application when the user had a respective known intent; 
 determine that the respective labeled subset was not entered by other users into the real-time communication application at a threshold frequency when the other users had the respective known intent; and 
 based on the respective labeled subset not being entered by other users into the real-time communication application at the threshold frequency when the other users had the respective known intent when the other users had the respective known intent, assign the respective labeled subset as corresponding to the respective known intent for the user; 
 
 in response to determining that the first subset corresponds to the first labeled subset in the plurality of labeled subsets, determining the user has a first known intent corresponding to the first labeled subset; and 
 generating for display, on the user interface, a first digital asset corresponding to the first known intent, wherein the first digital asset comprises a textual description related to the first known intent. 
   
     
     
         2 . A method for generating digital asset recommendations using artificial intelligence models in data-sparse environments, the method comprising:
 receiving, at a user interface, a first action, wherein the first action comprises a phrase entered by a user;   determining a first subset of the phrase;   determining whether the first subset corresponds to a first labeled subset in a plurality of labeled subsets, wherein each labeled subset of the plurality of labeled subsets corresponds to one of respective known intents of the user, and wherein the plurality of labeled subsets is based on an artificial intelligence model that is trained to:
 determine that a respective labeled subset of the plurality of labeled subsets was used when the user had a respective known intent; 
 determine that the respective labeled subset was not used at a threshold frequency by other users; and 
 based on the respective labeled subset not being used at the threshold frequency by other users, assign the respective labeled subset as corresponding to the respective known intent for the user; 
   in response to determining that the first subset corresponds to the first labeled subset in the plurality of labeled subsets, determining the user has a first known intent corresponding to the first labeled subset; and   generating for display, on the user interface, a first digital asset corresponding to the first known intent.   
     
     
         3 . The method of  claim 2 , wherein determining whether the first subset corresponds to the first labeled subset in the plurality of labeled subsets further comprises:
 parsing the phrase to identify a plurality of subsets; and   comparing each of the plurality of subsets to the first labeled subset.   
     
     
         4 . The method of  claim 2 , further comprising:
 receiving, at the user interface, a second action, wherein the second action comprises an additional phrase entered by a user;   determining a second subset of the additional phrase;   comparing the second subset to a user-selected intent identifier; and   in response to comparing the second subset to the user-selected intent identifier, determining that the user has a second intent corresponding to the user-selected intent identifier.   
     
     
         5 . The method of  claim 4 , further comprising:
 receiving, at the user interface, a third action, wherein the third action comprises a user request to create the user-selected intent identifier;   receiving a fourth action comprising a user input of a user-selected intent; and   receiving a fifth action comprising a user input of the user-selected intent identifier.   
     
     
         6 . The method of  claim 2 , wherein determining that the respective labeled subset of the plurality of labeled subsets was used when the user had the respective known intent further comprises:
 determining that the user had the respective known intent during a previous device session;   retrieving a previous phrase that was entered by the user during the previous device session; and   parsing the previous phrase to identify the respective labeled subset.   
     
     
         7 . The method of  claim 2 , wherein determining that the respective labeled subset of the plurality of labeled subsets was used when the user had the respective known intent further comprises:
 determining an input frequency of the respective labeled subset appearing in phrases entered by the user during previous device sessions during which the user had the respective known intent; and   comparing the input frequency to a threshold input frequency.   
     
     
         8 . The method of  claim 2 , wherein determining that the respective labeled subset was not used at the threshold frequency by other users further comprises:
 retrieving a plurality of previous phrases entered by the other users during previous device sessions;   parsing the plurality of previous phrases to identify an average frequency at which the respective labeled subset appeared in the plurality of previous phrases; and   comparing the average frequency to the threshold frequency.   
     
     
         9 . The method of  claim 2 , wherein determining that the respective labeled subset of the plurality of labeled subsets was used when the user had the respective known intent further comprises:
 generating for display, on the user interface during a previous device session, a textual description of the respective known intent; and   receiving a user confirmation of the respective known intent.   
     
     
         10 . The method of  claim 9 , wherein determining that the respective labeled subset of the plurality of labeled subsets was used when the user had the respective known intent further comprises:
 determining a first time, wherein the first time corresponds to a delivery time of the textual description;   determining a second time, wherein the second time corresponds to a receipt time of the textual description;   determining a length of time between the first time and the second time; and   comparing the length of time to a threshold length of time to determine that the respective labeled subset of the plurality of labeled subsets was used when the user had the respective known intent.   
     
     
         11 . The method of  claim 9 , wherein determining that the respective labeled subset of the plurality of labeled subsets was used when the user had the respective known intent further comprises:
 determining a first input number, wherein the first input number corresponds to a total number of inputs from the user at a receipt time of the textual description;   determining a second input number, wherein the second input number corresponds to the total number of inputs from the user at an end time of a current device session;   determining a difference between the first input number and the second input number; and   comparing the difference to a threshold difference to determine that the respective labeled subset of the plurality of labeled subsets was used when the user had the respective known intent.   
     
     
         12 . The method of  claim 2 , wherein generating for display, on the user interface, the first digital asset corresponding to the first known intent further comprises:
 retrieving a plurality of available digital assets;   determining respective known intents corresponding to each of the plurality of available digital assets; and   filtering the plurality of available digital assets based on the respective known intents corresponding to the first known intent.   
     
     
         13 . The method of  claim 2 , wherein generating for display, on the user interface, the first digital asset corresponding to the first known intent further comprises:
 retrieving a plurality of delivery formats corresponding to the first digital asset;   retrieving a user profile corresponding to the user; and   determining a delivery format of the plurality of delivery formats based on the user profile.   
     
     
         14 . The method of  claim 2 , wherein determining the user has the first known intent corresponding to the first labeled subset further comprises:
 determining a probability that the user has the first known intent; and   comparing the probability to a threshold probability.   
     
     
         15 . The method of  claim 2 , further comprising:
 determining a usage frequency of the respective labeled subset, wherein the usage frequency is specific to the user; and   determining the threshold frequency to apply to the respective labeled subset based on the usage frequency.   
     
     
         16 . A non-transitory, computer-readable medium comprising instructions recorded thereon that when executed by one or more processors causes operations comprising:
 receiving, at a user interface, a first action, wherein the first action comprises a phrase entered by a user;   determining a first subset of the phrase;   determining whether the first subset corresponds to a first labeled subset in a plurality of labeled subsets, wherein each labeled subset of the plurality of labeled subsets corresponds to one of respective known intents of the user, and wherein the plurality of labeled subsets is based on an artificial intelligence model that is trained to:
 determine that a respective labeled subset of the plurality of labeled subsets was used when the user had a respective known intent; 
 determine that the respective labeled subset was not used at a threshold frequency by other users; and 
 based on the respective labeled subset not being used at the threshold frequency by other users, assign the respective labeled subset as corresponding to the respective known intent for the user; 
   in response to determining that the first subset corresponds to the first labeled subset in the plurality of labeled subsets, determining the user has a first known intent corresponding to the first labeled subset; and   generating for display, on the user interface, a first digital asset corresponding to the first known intent.   
     
     
         17 . The non-transitory, computer-readable medium of  claim 16 , determining whether the first subset corresponds to the first labeled subset in the plurality of labeled subsets further comprises:
 parsing the phrase to identify a plurality of subsets; and   comparing each of the plurality of subsets to the first labeled subset.   
     
     
         18 . The non-transitory, computer-readable medium of  claim 16 , further comprising:
 receiving, at the user interface, a second action, wherein the second action comprises an additional phrase entered by a user;   determining a second subset of the additional phrase;   comparing the second subset to a user-selected intent identifier; and   in response to comparing the second subset to the user-selected intent identifier, determining that the user has a second intent corresponding to the user-selected intent identifier.   
     
     
         19 . The non-transitory, computer-readable medium of  claim 18 , further comprising:
 receiving, at the user interface, a third action, wherein the third action comprises a user request to create the user-selected intent identifier;   receiving a fourth action comprising a user input of a user-selected intent; and   receiving a fifth action comprising a user input of the user-selected intent identifier.   
     
     
         20 . The non-transitory, computer-readable medium of  claim 16 , wherein determining that the respective labeled subset of the plurality of labeled subsets was used when the user had the respective known intent further comprises:
 determining that the user had the respective known intent during a previous device session;   retrieving a previous phrase that was entered by the user during the previous device session; and   parsing the previous phrase to identify the respective labeled subset.